A Data-Driven Approach to Assessing and Improving Stochastic Residential Load Modeling for District-Level Simulations and PV Integration

This paper presents an assessment and improvement of stochastic load modeling for district-level analyses with integration of photovoltaic panels (PV), by comparison with measurement data. Stochastic load profiles for individual households were produced using the bottom-up `Stochastic Residential Occupancy Behavior' (StROBe) model. The self-consumption of households with PV installations and the district-level peak demand are examined as properties relevant for the estimation of PV hosting capacity and accompanying grid-related problems. The comparison shows that while the synthetic profiles produce reasonable estimates of simultaneity and summer peak demand, they insufficiently represent the seasonal variations. In addition, self-consumption is overestimated by the model. The observed discrepancies can be traced back to inaccurate modeling of the peak timing and seasonal variation in individual peak load and simultaneity. Furthermore, vacant homes in the measured data are found to contribute significantly to discrepancies in holiday periods. Adjusting the stochastic modeling to account for these vacant homes results in improved performance of the model. This research demonstrates that harvesting the full potential of bottom-up stochastic load modeling would require more up-to-date information on residential electricity use patterns.